def execute(self, experiment: Experiment) -> ExperimentPerformance:
"""
Execute K-fold cross-validation with separate test set with the given classifier on the given data set using the
given parameters.
PARAMETERS
----------
experiment : Experiment
Experiment to be run.
RETURNS
-------
ExperimentPerformance
An ExperimentPerformance instance.
"""
result = ExperimentPerformance()
instanceList = experiment.getDataSet().getInstanceList()
partition = Partition(instanceList, 0.25,
experiment.getParameter().getSeed(), True)
crossValidation = KFoldCrossValidation(
partition.get(1).getInstances(), self.K,
experiment.getParameter().getSeed())
self.runExperiment(experiment.getClassifier(),
experiment.getParameter(), result, crossValidation,
partition.get(0))
return result
def train(self, trainSet: InstanceList, parameters: Parameter):
"""
Training algorithm for the linear discriminant analysis classifier (Introduction to Machine Learning, Alpaydin,
2015).
PARAMETERS
----------
trainSet : InstanceList
Training data given to the algorithm.
parameters : Parameter
Parameter of the Lda algorithm.
"""
w0 = {}
w = {}
priorDistribution = trainSet.classDistribution()
classLists = Partition(trainSet)
covariance = Matrix(trainSet.get(0).continuousAttributeSize(), trainSet.get(0).continuousAttributeSize())
for i in range(classLists.size()):
averageVector = Vector(classLists.get(i).continuousAverage())
classCovariance = classLists.get(i).covariance(averageVector)
classCovariance.multiplyWithConstant(classLists.get(i).size() - 1)
covariance.add(classCovariance)
covariance.divideByConstant(trainSet.size() - classLists.size())
covariance.inverse()
for i in range(classLists.size()):
Ci = classLists.get(i).getClassLabel()
averageVector = Vector(classLists.get(i).continuousAverage())
wi = covariance.multiplyWithVectorFromRight(averageVector)
w[Ci] = wi
w0i = -0.5 * wi.dotProduct(averageVector) + math.log(priorDistribution.getProbability(Ci))
w0[Ci] = w0i
self.model = LdaModel(priorDistribution, w, w0)
def train(self, trainSet: InstanceList, parameters: KMeansParameter):
priorDistribution = trainSet.classDistribution()
classMeans = InstanceList()
classLists = Partition(trainSet)
for i in range(classLists.size()):
classMeans.add(classLists.get(i).average())
self.model = KMeansModel(priorDistribution, classMeans,
parameters.getDistanceMetric())
def train(self, trainSet: InstanceList, parameters: DeepNetworkParameter):
"""
Training algorithm for deep network classifier.
PARAMETERS
----------
trainSet : InstanceList
Training data given to the algorithm.
parameters : DeepNetworkParameter
Parameters of the deep network algorithm. crossValidationRatio and seed are used as parameters.
"""
partition = Partition(trainSet, parameters.getCrossValidationRatio(), parameters.getSeed(), True)
self.model = DeepNetworkModel(partition.get(1), partition.get(0), parameters)
def execute(self, experiment: Experiment) -> ExperimentPerformance:
result = ExperimentPerformance()
for j in range(self.M):
instanceList = experiment.getDataSet().getInstanceList()
partition = Partition(instanceList, 0.25,
experiment.getParameter().getSeed(), True)
crossValidation = StratifiedKFoldCrossValidation(
Partition(partition.get(1)).getLists(), self.K,
experiment.getParameter().getSeed())
self.runExperiment(experiment.getClassifier(),
experiment.getParameter(), result,
crossValidation, partition.get(0))
return result
def train(self, trainSet: InstanceList, parameters: MultiLayerPerceptronParameter):
"""
Training algorithm for the multilayer perceptron algorithm. 20 percent of the data is separated as
cross-validation data used for selecting the best weights. 80 percent of the data is used for training the
multilayer perceptron with gradient descent.
PARAMETERS
----------
trainSet : InstanceList
Training data given to the algorithm
parameters : MultiLayerPerceptronParameter
Parameters of the multilayer perceptron.
"""
partition = Partition(trainSet, parameters.getCrossValidationRatio(), parameters.getSeed(), True)
self.model = MultiLayerPerceptronModel(partition.get(1), partition.get(0), parameters)
def train(self, trainSet: InstanceList,
parameters: MultiLayerPerceptronParameter):
"""
Training algorithm for auto encoders. An auto encoder is a neural network which attempts to replicate its input
at its output.
PARAMETERS
----------
trainSet : InstanceList
Training data given to the algorithm.
parameters : MultiLayerPerceptronParameter
Parameters of the auto encoder.
"""
partition = Partition(trainSet, 0.2, parameters.getSeed(), True)
self.model = AutoEncoderModel(partition.get(1), partition.get(0),
parameters)
def getClassInstances(self) -> list:
"""
Returns instances of the items at the list of instance lists from the partitions.
RETURNS
-------
list
Instances of the items at the list of instance lists from the partitions.
"""
return Partition(self.__instances).getLists()
def train(self, trainSet: InstanceList, parameters: C45Parameter):
"""
Training algorithm for C4.5 univariate decision tree classifier. 20 percent of the data are left aside for
pruning 80 percent of the data is used for constructing the tree.
PARAMETERS
----------
trainSet : InstanceList
Training data given to the algorithm.
parameters: C45Parameter
Parameter of the C45 algorithm.
"""
if parameters.isPrune():
partition = Partition(trainSet,
parameters.getCrossValidationRatio(),
parameters.getSeed(), True)
tree = DecisionTree(DecisionNode(partition.get(1)))
tree.prune(partition.get(0))
else:
tree = DecisionTree(DecisionNode(trainSet))
self.model = tree
def train(self, trainSet: InstanceList, parameters: Parameter):
"""
Training algorithm for the quadratic discriminant analysis classifier (Introduction to Machine Learning,
Alpaydin, 2015).
PARAMETERS
----------
trainSet : InstanceList
Training data given to the algorithm.
"""
w0 = {}
w = {}
W = {}
classLists = Partition(trainSet)
priorDistribution = trainSet.classDistribution()
for i in range(classLists.size()):
Ci = classLists.get(i).getClassLabel()
averageVector = Vector(classLists.get(i).continuousAverage())
classCovariance = classLists.get(i).covariance(averageVector)
determinant = classCovariance.determinant()
classCovariance.inverse()
Wi = deepcopy(classCovariance)
Wi.multiplyWithConstant(-0.5)
W[Ci] = Wi
wi = classCovariance.multiplyWithVectorFromLeft(averageVector)
w[Ci] = wi
w0i = -0.5 * (wi.dotProduct(averageVector) + math.log(determinant)) + math.log(priorDistribution.
getProbability(Ci))
w0[Ci] = w0i
self.model = QdaModel(priorDistribution, W, w, w0)